Land Use Land Cover (LULC) Change Detection Using Geospatial Technique in Anbessa Forest, Benishangul-Gumuz Region, Ethiopia

LULC analysis using satellite images for detecting the changes across a given landscape is a very crucial tool for understanding the nexus between forest ecosystems and human activities. LULC pattern of Anbessa forest had undergone fast changes over the last 30 years, but no research measured the level of the changes. The present study was aimed at detecting the LULC change pattern of Anbessa forest using data from satellite images between 1989 and 2019. Methods We examined the LULC changes of Anbessa forest using satellite image data over the period of 1989–2019 using geospatial technique. The results show a 29% and 18% decrease in land area under dense and open forests respectively over a period of 30 years. Conversely there is 32% and 25.6% increase in the land under agricultural land and settlement areas respectively. A relatively small average decrease of 8% in shrub land was found although there was a decrease from 1989 to 2004 and an increase from 2004 to 2019.


Abstract
Background LULC analysis using satellite images for detecting the changes across a given landscape is a very crucial tool for understanding the nexus between forest ecosystems and human activities. LULC pattern of Anbessa forest had undergone fast changes over the last 30 years, but no research measured the level of the changes.
The present study was aimed at detecting the LULC change pattern of Anbessa forest using data from satellite images between 1989 and 2019.

Methods
We examined the LULC changes of Anbessa forest using satellite image data over the period of 1989-2019 using geospatial technique.

Results
The results show a 29% and 18% decrease in land area under dense and open forests respectively over a period of 30 years. Conversely there is 32% and 25.6% increase in the land under agricultural land and settlement areas respectively. A relatively small average decrease of 8% in shrub land was found although there was a decrease from 1989 to 2004 and an increase from 2004 to 2019.

Conclusion
The fact that there is a decrease in natural and open forests and an increase in agricultural and settlement areas implies there had been fast degradation of natural forests of Anbessa forest due to human activities.
Thus, there should be an intervention project that ensures the sustainability of the forest.

Background
Globally only very few landscapes remain without signi cant alteration by human activities (Yang, 2001).
Most landscapes have experienced land use land cover (LULC) change, which has become a major topic in, sustainable management of natural resources (Kadıoğullarıh, 2013) and, sustainable development (Hu et al., 2018) since recently. LULC change has had considerable effect on forest ecosystem, biodiversity, soil conservation and global climate (Johnson et al., 1997;Xu et al. 2007;Karnieli et al., 2008).Human activities are the main causes of LULC changes through ever increasing population pressure and the resulting demand for agricultural land, wood, charcoal and rewood production from forests, overgrazing and indiscriminate cutting of trees (Kennedy &Spies2004;Wakeeletal., 2005;Cayuelaetal., 2006).So, understanding the spatial extent and distribution of LULC change is vital to the study of environmental changes at various levels (Ojima et al. 1994).Accordingly, environmental analysts recognized that LULC analysis is a fundamental tool for assessing environmental and ecological consequences of human activities (Yang, 2001;Flamenco-Sandoval et al., 2007).But, what is LULC change?
LULC change refers to conversion of one type of LULC to another (Turner & Meyer, 1994). It also refers to human modi cation of the terrestrial surface of the Earth, and re ects the role of human activities on natural resources and the environment (Shah & Sharma, 2015). Whereas land cover denotes the biophysical attributes of the Earth's surface, land use refers to the human purpose or intent applied to these attributes . Forest cover is one of the LULC types for which both attributes are considered in detecting the changes in Anbessa forest, which is dominantly wood grassland at present (Tamene, 2016).
The structure of forest ecosystem refers to the spatial characteristics of ecosystem including the size, shape, composition, and spatial arrangement (Kadıoğulları, 2013). The environmental and economic functions of forest ecosystems therefore, depend on these characteristics, are affected by LULC changes and, need proper management (Shah & Sharma, 2015). Thus, the sustainability of the services of forest ecosystems depends on their health and stability.
Forest ecosystems provide not only conducive environment for human survival but also home for numerous animals. Forests harbor two thirds of all the terrestrial animal and plant species (World Bank, 2004). Their environmental and economic functions such as timber, fuel wood, fodder, water and soil protection, carbon sequestration, oxygen production, recreation, aesthetics, biodiversity, and habitat for wildlife species are described in vast literature (Köchli &Brang 2005;Keleş et al. 2008;Başkent et al. 2008). Specially, their economic functions such as contribution to food security in general (Shriar, 2002) and being a major source of wild foods in particular are paramount (Guyu & Muluneh, 2015). However, forests of Ethiopia including Anbessa forest have experienced LULC changes due to mainly human activities (McKee, 2007;Mathewos, 2019).
Deforestation is an important cause of LULC elsewhere in the world (Lambin et al., 2001) and also in forest areas of Ethiopia (McKee, 2007;Mathewos, 2019). The major drivers of deforestation in Ethiopia are settlements, agriculture (both small scale and commercial), extraction of construction materials, grazing, and rewood and charcoal collection (McKee, 2007). In the present study, deforestation is thus de ned as high forest being converted into other land-use types. Forest structure and composition are essential ecosystem characteristics (Quesada & Kuuluvainen, 2020) which are disturbed by deforestation. The regular and periodic assessments of forest cover change in tropical regions are therefore carried out to recognize previous patterns, assist proper planning and predict future trends (Shah and Sharma, 2015).
In Ethiopia, forest losses of 140,000 hectares each year are driven by conversion into agricultural lands, and unsustainable forest management, underpinned by poor governance, uncertain land tenure and a rapidly growing population (Lawrence et al., 2010). The average annual deforestation rate is 1% which is high compared to other Sub-Saharan African countries (0.6%) (Tamene, 2016). Anbessa forest, one of the popular high forests of Ethiopia, has been exposed to deforestation over the last three to four decades. But, there is no research that shows the level of deforestation, and the type and nature of LULC changes that have so far occurred in the forest. A previous study examined plant diversity in the forest (Tamene, 2016).Therefore, understanding the level of LULC changes in quantitative and qualitative terms is necessary to design appropriate intervention strategies. Therefore, the aim of the present study was to detect the level of LULC change that had occurred in Anbessa forest over the last 30 years using geospatial techniques so as to understand the spatial and temporal dynamics of the size and pattern of forest cover changes.

Study Area
Anbessa forest is located in Bambasi district, Benishangul-Gumuz region, Ethiopia. Astronomical location of the forest ranges from 9 0 53 24.3′′ N -9 0 55 40.8′′ N and from 34 0 39 09.0′′ E − 34 0 50 55.3′′ E. Since recently, the estimated total area of the forest has become 15,072 ha (Tamene, 2016). However, originally it had total area of about 50,213 ha (Fig. 1). The topography of Anbessa forest is very at except for few hills in the western part of the forest near the main asphalt road from Addis Ababa (Capital city of Ethiopia) to Asossa (capital town of Benishangul-Gumuz region). Its elevation ranges from 1292 m to 1563 m a.s.l. with the highest peak in the western side and the lowest in the eastern side.
As part of the western lowlands of Ethiopia, Anbessa forest is characterized by unimodal rainfall distribution with the rainy season extending from March to November (Tamene, 2016). Its average annual rainfall is 1381.42 mm and the mean monthly temperature is 28.37 0 C. Hydrologically, Anbessa forest is found in the Blue Nile river basin. There are a number of big and small rives which are tributaries of Blue Nile, such as Dabus, Afa, Selga, Shosha, Mutsa, Ni ro, Abakidi, Eshama, Chilonya and many small streams which pass through or near by the forest (Herrmann et al., 2007).
There is no signi cant variability in the vegetation type due to low level of variations in its altitude. In general, Anbessa forest can be characterized as Combretum-Terminalia woodland. The forest is dominated by Oxytenanthera abyssinica (lowland bamboo) which stands with scattered Combretum -Terminalia Woodland vegetation. The rest of the forest is at wooded grass land with very small slope variation.
Signi cant level of deforestation on Anbessa forest has started since the implementation of resettlements schemes in 1980s which brought settler Amhara ethnic group from northern Ethiopia to the area. At present, the forest is surrounded by 10 large villages of this type namely: Amba 16, Jematsa, Garabiche Welega, Sonka, Village 44, and Village 47 (Fig. 1). Since then, theforest has been under serious pressure from deforestation due to the increasing number of population around it and the resulting demand for arable agriculture, grazing land and production of charcoal, rewood and timber. As a result, it is a typical area that faces many ecological problems (Tamene, 2016).

Data
Forest resource maps had been traditionally prepared from forest inventories involving aerial photography and eldwork. However, the modern technology, Geographic Information System (GIS) technique and Remote Sensing (RS) from Satellite platforms offer an alternative and economic tool for forest mapping (Shah & Sharma, 2015). In the present study too, LULC types were mapped as the spatial database from satellite images obtained from three Landsat series.
The images over the period of thirty years were derived from Landsat Thematic Mapper (TM), Landsat Enhanced Thematic Mapper Plus (ETM+) and Landsat 8 Operational Land Imager/Thermal Infrared Sensor (OLI/TIR) for 1989, 2004 and 2019 respectively. The year 1989 is chosen as the base because settlement programs started in the rst half of the year in the area and satellite data are available for the study area since then. The Landsat satellite images for these years were acquired from the U.S Geological Survey (USGS) via Google Earth to evaluate the forest cover change over the period of 1989 to 2019. The wave length of the Landsat TM sensor ranges from the visible to the thermal infrared portion of the electromagnetic spectrum and has a spatial resolution of 30 meter. After six bands of the TM (excluding thermal band) were considered for layer stacking, TM band 4, 3 and 2 were combined to make conventional false color composite image.
The Landsat ETM + was introduced in Landsat 7 and its data covered the visible, near-infrared, shortwave, and thermal infrared spectral bands of the electromagnetic spectrum. ETM+, which improved the version of TM sensor, has thermal band with an improved spatial resolution of 60metercompare to the TM′s 120meter spatial resolution. The ETM + also contains 15 m panchromatic band. After six bands of the ETM+ (excluding thermal band) were considered for layer stacking, ETM + band 4, 3 and 2 were combined to make conventional false color composite images.
NASA successfully launched the Landsat Data Continuity Mission on 11 February, 2013. The satellite was renamed Landsat 8 and operation has been transferred to the USGS. Data collected since April 11, 2013 by the OLI and TIRS on board Landsat 8 are available for download. Of its 11 bands, only those in the very shortest wavelengths (bands 1-4 and 8) sense visible light whereas all the others are in parts of the spectrum those we cannot see. The true-color view from Landsat is less than half of what it sees. As a result, the images need to be contrasted and enhanced (stretched). Following this recommendation, histogram equalization was run to enhance the image and a good result was found. Over seven bands of the Landsat 8 (excluding thermal band) were considered for layer stacking, and band 5, 4 and 3 were combined to make conventional false color composite images with a spatial resolution of 30 meters.

Image Processing and Presentation
The LULC change was detected using post classi cation cross-tabulation approach in ARC-GIS and ERDAS.
The images were processed in three phases: pre-processing, image classi cation, and post-classi cation and change detection.
In the rst phase, radiometric and atmospheric corrections were performed to correct atmospheric conditions from sensors' scanning errors as well as distortions from solar angle and sensors' angle in the Landsat images. Then, all images from each study year were clipped to match with the study area and buffer zone boundaries.
In the second phase, image classi cation and land cover change detection were performed. Supervised classi cation method was applied to three Landsat images (1989,2004 and 2019) using the Maximum Likelihood Classi cation method (MLC) -a well-known parametric classi er for supervised classi cation (Otukei & Blaschke, 2010).
In the third phase, post classi cation process and change detection was performed to assess the changes through the entire land cover during the study period for both the forest covered area and degraded zone.
Land cover maps were produced for each year for which satellite images were used. In this process, the forest classes were identi ed and given names as dense forest, open forest, shrub land, agricultural land and settlement. The results are presented qualitatively in maps and quantitatively in tables, and are synthesized by drawing conclusion.

Spatial Pattern of Anbessa Forest Cover in 1989, the Reference Year
The1989 LULC map after the Supervised Classi cation yielded the spatial extent of different land cover classes (Fig. 2).There was a clear association between settlement and agricultural lands in the forest. Spatially, these land use types largely found in northwestern and north central with close proximity to dense forests and open forests. The eastern half of Anbessa forest area that extends from North through Northeast and Southeast to South was occupied by settlement and agricultural lands.
On the other hand, dense forest cover was found largely in western half of the forest area although signi cant distribution of this type of forest cover extended from southern via southeastern and central to northern parts. Open forest covers were also found in north, north central, northwest, central, northeastern, eastern and southeastern parts of the forest. Open forests are generally associated with the shrub lands ( Figure, 2). In general, dense and open forests covered more area than other types of LULC. This implies that deforestation was very smaller than the next decades. The amount of LULC types is also measured quantitatively for clear detection. The dense forests covered the highest part of Anbessa forest area accounting for 15,884 ha (35.63%) followed by open forest with 14,177.2 ha (31.80%) and shrub land with 12,184.7 ha (27.33%). The smallest LULC types were agricultural land and settlements constituting about 1,724 ha (3.9%) and 599 (1.34%) respectively (Table 1).

Spatial Pattern of LULC Classes of 2004 and its Change from 1989
The supervised classi cation procedure applied to the 2004 image shows the largest land cover map for agricultural land and settlement as compared to other LULC classes (Fig. 3)

Spatial Pattern of LULC Classes of 2019 and the Change from 2004
Analyzing the pattern of LULC classes of the 2019, mapped from sattelite images,is crutial step to compare with maps of 1989 and 2004.The result showed that the change in LULC had continued to happen until 2019 (Fig. 4). The northeastern part of the forest area was almost entirely conversted into agricultural land. Large part of northern, northcentral, central and southeastern part of the area was dominantly occupied by agricultural land. Settlements were also distributed almost all over the area of Anbessa forest. Most part of Anbessa forest had also gradually been changed into shrub lands.
In contrast to agricultural, settlement and shrub lands, dense forest and open forest covers had decreased from 2004 to 2019. Dense forests likely divided Anbessa forest into two running from North through central part to Southwest. These forests were also found in central southeastern part of the forest area. Open forests were found in northwestern and western as well as northeastern fringes of the forest area. In addition, these classes of forest area were distributed throughout Anbessa forest (Fig. 4). The spatial map of the LULC of the forest during the reference year (1989) was in line with the general observation where dense forests had the largest area (Fig. 2). There was a clear association between settlement and agricultural lands in the forest. This goes in line with the fact that human cultural interventions are the main causes of environmental changes (Yang, 2001;Johnson et al., 1997;Xu et al., 2007). Thus, it is natural that LULC changes start when human beings to encroach forests through their activities. This was true in Anbessa forest where the spatial and temporal pattern of changes had started since the1984/85 resettlement scheme due to drought in the northern part of Ethiopia brought many people to this area (Tamene, 2016). Since then, the natural forest cover had shown a signi cant decrease giving ways to agricultural and settlement areas.  (Xu et al. 2007;Karnieli et al., 2008;Hu et al., 2018;Mathewos, 2019). This clearly implies that there was low level of interference in the forest ecosystem via human activities such as agriculture until 1989. However, the subsequent years were accompanied by increased human interference followed by a fast decline in the natural forest co The above facts are proved by observing the rate of changes in the LULC (Table 5) changes in agricultural and settlement areas over the same period of time (Fig. 4). So we can conclude that Anbessa forest had experienced a change from natural landscape to cultural landscape over the last 30 years.  (Table 4).The negative changes in the dense and open forests and positive changes in the agricultural and settlement lands (Table 5) were also our prior expectation. This goes in line with the general observation that human activities such as deforestation for settlements and agriculture are the main causes of LULC changes (Turner and Meyer, 1994;Yang, 2001;Shah and Sharma, 2015).This obviously shows that human cultural practices play signi cant role in LULC changes.

Conclusion
LUCC analysis is a crucial instrument that provides information on the environmental changes at any scale.
The ndings of the analysis of LULC are useful for decisions makers to work towards sustainable environmental management and development. In this study, we used a long term series of satellite images (1989, 2004 and 2019)to analyze the changes that had occurred in Anbessa forest. Our ndings showed that there had been fast changes in LULC between 1989 and 2019. The transition, from dense and open forests to agricultural and settlement areas, was considerable. This was followed by major decrease in the areas of natural dense and open forests and an increase in agricultural and settlement areas over 30 years. In general, although the average change in all types of LULC was negative, different directions of changes were observed. That is, there was a negative change in the land cover under dense and open forests and a positive change in land cover under agriculture and settlement while shrub lands experienced an initial decrease followed by an increase from 1989 to 2004 and to 2019 respectively. Thus, we concluded that Anbessa forest had faced critical environmental changes mainly due to encroachments through agriculture and settlements. We recommend an intervention projects that can avert the current trend of Anbessa forest degradation. More importantly, we also suggest further research into the root causes for the reliance of the